2.1 Trial design and participants

This RCT was conducted at the University Medical Centers Hamburg-Eppendorf and Leipzig outpatient psycho-oncological counselling centres, approved by local ethic committees (PV4435; 143–14–14042 014), registered on ClinicalTrials.gov (NCT02051660) and pilot-tested.¹⁵

Eligible patients were ≥18 years, fluent in German, and had a malignant solid tumor (UICC III or IV). Patients were included with a score of ≥9 on the Patient Health Questionnaire (PHQ-9)¹⁶ and/or ≥5 on the Distress Thermometer (DT).¹⁷ Patients were ineligible if they had severe cognitive impairment (≤20 on Short Orientation-Memory-Concentration Test, SOMC),¹⁸ a high level of functional impairments (Karnofsky score ≤70),¹⁹ parallel psychotherapeutic treatment or if they reported suicidal ideation.

We recruited from August 2013 to January 2017 (last finished treatment in June 2017). All patients provided written informed consent. We enrolled 206 instead of originally aimed 262 patients²⁰ since the completion rate (CALM: 69 t₂ assessments; SPI: 56 t₂ assessments; Figure 1) was in fact higher than the estimated completion rate (50 t₂ assessments required in each group²⁰). Participants remained in the study until they reached the maximum number of sessions, unless they withdrew their consent or died.

2.2 Randomization and masking

Randomization and sequence generation was conducted by the Institute of Medical Biometry and Epidemiology at the University Medical Center Hamburg-Eppendorf, which was not involved in patient recruitment or data collection. We used a computer random number generator for sequence generation. Randomization was stratified for trial site and sex. No study staff member had access to the randomization list (concealed allocation sequence). To prevent performance bias, patients were not informed about treatment condition, neither by study staff members nor by their therapists (single-blind).

2.3 Trial interventions

The CALM manual⁹ covers four domains: (a) SC: symptom management and communication with health care providers; (b) CSR: changes in self and relations with others, (c) SMP: spirituality, sense of meaning and purpose, (d) FHM: preparing for the future, sustaining hope and facing mortality. We translated the original CALM manual⁹ resulting in a German language research version.

SPI is a non-manualized psychotherapeutic counselling intervention that has been routinely conducted in both study centres for several years. SPI therapists were instructed to perform their best usual practice. SPI helps patients dealing with distressing emotions by promoting adaptive coping with advanced cancer.²¹ Therapeutic techniques are derived from cognitive-behavioural techniques and/or psychodynamic treatment, including psychoeducation and emotion regulation.

2.4 Trial conditions

CALM (13 therapists) and SPI (14 therapists) was conducted by licensed psychotherapists with comparable treatment experience (CALM: M = 15.1 years; SD = 8.3, range 8-30; SPI: M = 12.8 years; SD = 9.1, range: 2-30) randomized to both arms. CALM therapists were M = 41.6 years of age (SD = 9.5, range: 32-60) and SPI therapists M = 43.7 years of age (SD = 11.9, range: 29-64). Sixty-nine percent (n = 9) of CALM therapists were female as were 71% (n = 10) of SPI therapists.

CALM therapists were trained in a two-day workshop held by CALM-experienced psycho-oncologists (A.M., D.E., F.S.-K.) and received supervision every 4 weeks by CALM-experienced supervisors. SPI therapists were blinded to the content of CALM and received the same amount of supervision by non-CALM-trained supervisors.

Within 6 months, CALM and SPI patients received up to six treatment sessions (á 50 minutes) and two optional booster sessions. Three sessions were the minimum dose of therapy, patients with at least six sessions were declared as treatment completers. If patients expressed their need for further treatment, therapy was continued.

2.5 Treatment integrity and treatment differentiation

All CALM and SPI treatments were audio recorded. Therapists' adherence to the CALM manual and treatment differentiation was assessed in a random sample of 150 audio recordings of 50 patients (25 CALM, 25 SPI). Two trained raters with a master in clinical psychology used the Treatment Integrity Rating Scale¹³ to assess four CALM domains (see Supporting Information): SC, CSR, SMP and FHM.⁹ The extent of therapists' adherence was estimated on a three-point Likert scale with “1 = adherent to some extent” for an acceptable, but not satisfactory implementation, “2 = adherent to a sufficient extent” for a satisfactory implementation, “3 = very adherent” for an excellent implementation of a CALM technique as described in the CALM manual, (Koranyi et al., in revision). Raters were blinded to treatment condition. Treatment integrity scores represent the consensus version of both raters.

2.6 Outcome measures

Outcome assessments took place at baseline (t₀), 3 months after randomization (t₁) and 6 months (t₂) after randomization. The primary outcome was depression severity assessed with BDI-II²² and PHQ-9¹⁶ at t₂. Both self-report measures are reliable and validated instruments.^{23, 24}

Secondary outcomes were psychological distress (DT¹⁷), quality of life (Quality of Life at the End of Life Cancer, QUAL-EC²⁵), anxiety (Generalized Anxiety Disorder Questionnaire, GAD-7²⁶), demoralization (Demoralization Scale, DS²⁷), fatigue (Brief Fatigue Inventory, BFI²⁸), symptom burden (Memorial Symptom Assessment Scale, MSAS-SF²⁹), spiritual well-being (Functional Assessment of Chronic Illness Therapy Spiritual Well-Being Scale, FACIT-Sp³⁰), death anxiety (Death and Dying Distress Scale, DADDS³¹), posttraumatic growth (Posttraumatic Growth Inventory, PTGI³²), attachment insecurity (Experiences in Close Relationships Scale, ECR-M16³³), and couple communication (Couple Communication Scale, CCS³⁴).

2.7 Sample size calculation

The sample size was determined for testing the hypothesis that CALM is superior to SPI regarding the reduction of depressive symptoms. We aimed to detect a 3-points mean difference on the BDI-II between groups at t₂. Using an expected SD of 7.4,³⁵ this corresponds to a small to moderate standardized effect (Cohen's d of .405). Assuming a correlation of .70 between baseline and follow-up measures,³⁵ 50 participants per group are necessary to detect the targeted effect with an alpha of .05 and a power of .80.³⁶ Accounting for an expected completion rate of .60 and an expected intervention compliance rate of .80, this implies the necessity of recruiting 131 participants per group.³⁶

2.8 Statistical analysis

The main analyses were performed in the intention-to-treat (ITT) population. In contrast to the study protocol,²⁰ we used linear mixed modelling (assuming that data are missing at random, conditional on the information in the model) rather than analysis of covariance with last observation carried forward due to its clear superiority regarding dealing with missing data.³⁷ This decision was made prior to data analysis. The outcomes at t₀, t₁, and t₂ were modelled with intervention group, time point, trial site, sex, and age as fixed main effects, the interaction intervention group X time point to test group differences in the course of the outcome, and a random intercept to account for inter-individual differences.

We repeated all analyses using analysis of covariance with all available data without imputation, that is, with analyzing only patients who were not lost to follow-up at each measurement point. These models included intervention group, time point, site, sex, age, and baseline level of the respective outcome as fixed main effects.

We calculated standardized effect sizes (Cohen's d) for all group comparisons by dividing the model-based adjusted mean difference by the pooled observed SDs.³⁸ For testing the two primary outcomes (BDI-II and PHQ-9), we used a two-sided significance level (p) of .025 (corrected for multiple testing according to Bonferroni). For all other analyses, findings with P < .05 were considered statistically significant. As the level of significance was not adjusted for multiple testing of the secondary outcomes, P-values regarding any but the primary outcomes should be considered exploratory.

All analyses were performed with IBM SPSS Statistics 25 (IBM Corporation, Armonk NY).

3.1 Sample characteristics

A total of 2071 patients were assessed for eligibility and 123 of 329 eligible patients declined participation (37% non-participants) (Figure 1). Participants and non- participants did not differ in age (M__participants = 57.9, SD = 11.7; M__{non-participants} = 57.7, SD = 12.0) and sex (61.2% female participants and 63.1% female non-participants). Participants were more frequently diagnosed with gastrointestinal (60% vs 40%), lung (25% vs 15%), and breast cancer (33% vs 19%) (P < .001). Participants more frequently completed junior high school (30% vs 23%) and high school (50% vs 16%), non-participants more often received a college degree (40% vs 36%) (P < .05).

The 206 patients were randomly assigned to CALM (n = 99) or SPI (n = 107). The two groups did not differ in age, sex, education, or type of cancer (Table 1).

Both the average number of sessions (CALM: M = 5.2, SD = 3.0; SPI: M = 4.7, SD = 2.9) and the average length of treatment in weeks (CALM: M = 20.2, range 0-50; SPI: M = 19.5, range: 0-69) were distributed equally. Overall, 150 patients received a minimum dose of three therapy session (CALM: n = 73, SPI: n = 77) and 104 patients completed therapy with at least six therapy sessions (CALM: n = 54, SPI: n = 50). Attrition rates were at t₁ CALM 15%, SPI 24%, and at t₂ CALM 24% and SPI 40%.

3.2 Treatment integrity

CALM therapists showed low adherence scores (SC: M = 1.44; SD = 0.44; CSR: M = 1.46; SD = 0.44; SMP: M = 1.25; SD = 0.41; FHM: M = 1.38; SD = 0.52). SPI therapists showed also low adherence scores (SC: M = 1.42; SD = 0.52; CSR: M = 1.41; SD = 0.37; SMP: M = 1.32; SD = 0.61; FHM: M = 1.36; SD = 0.50).

CALM therapists did significantly apply more FHM techniques than SPI therapist (P < .05). No significant differences were observed on SC, CSR, and SMP dimensions.

3.3 Primary outcomes

At baseline, the groups did not differ significantly in the severity of depressive symptoms (BDI-II: M__CALM = 20.10, SD = 8.59; M__SPI = 18.20, SD = 8.83; PHQ-9: M__CALM = 11.41, SD = 5.16; M__SPI = 11.03, SD = 4.94). Although participants reported significantly less depressive symptoms at t₂ than at baseline (P < .001 for both primary outcomes), participants in the CALM and SPI group did not differ significantly in depression severity at t₂ (BDI-II: d = 0.06, 95%CI [−0.28; 0.35], P = .62; PHQ-9: d = 0.04, 95%CI [−0.21; 0.37], P = .998). These findings were confirmed in the analysis of available cases (Table 2).

3.4 Secondary outcomes

At baseline, the groups did not differ significantly in any of the secondary outcomes, except for attachment insecurity: CALM participants reported significantly higher attachment insecurity than SPI participants (ECR-M16: M__CALM = 52.5, SD = 16.1; M__SPI = 46.9, SD = 15.5). At t₁, CALM participants reported significantly higher demoralization (P = .01) and attachment insecurity (P < .05) than those in the SPI group. At t₂, there was no difference between the groups for any of the secondary outcomes (Table 3).

5.1 Study limitations

It was not possible to determine whether the decrease of depression was an intervention effect or a natural improvement over time. To answer this question, we would have required a treatment arm without psycho-oncological support (often indicated as care as usual in other trials) or a waiting list. However, in both study centres SPI is the care which is usually offered to patients with advanced cancer who report enhanced emotional distress. It would have been ethically inappropriate to withhold this routine psycho-oncological support from distressed patient. In addition, an improvement of depressive symptoms over time without treatment is rather unlikely since depression and quality of life in patients with advanced cancer worsen over time.^{13, 39, 40}

Further limitations are the low scores of treatment adherence in CALM therapists. We have chosen a health services approach and a non-manualized control intervention instead of a manualized SPI intervention, with its own training, supervision and fidelity measure. Future studies should test the hypothesis whether CALM is inferior to a manualized supportive or supportive-expressive psycho-oncological therapy.

Furthermore, the considerable rate of non-participants (37% of eligible patients) is leading to a decreased generalizability of results. The question arises as to what intervention formats are suitable for patients who feel too distressed for study participation (19% of non-participants). This should be the focus of future investigations.

5.2 Clinical implications

CALM therapy focuses on key aspects in psycho-oncological care for patients with advanced cancer. Psychotherapists who are inexperienced in psycho-oncological care should receive proper, fulsome training in psycho-oncology including education in validated, manualized supportive psychotherapeutic approaches for advanced cancer patients (eg, dignity therapy, meaning-centred psychotherapy and CALM therapy). This may be valuable for patients with a short time to address core elements in living with advanced disease and preparing for impending mortality.